A Comprehensive Survey on Parametric and Non-Parametric Machine Learning Approaches for Bus Arrival Time Prediction

Abstract

Public transportation plays a major role in modern society. Bus transportation is vitally important in public transportation, and it reduces more than 60% of private vehicle usage. It reduces fuel exhaustion and traffic burdens. Hence, promoting the usage of public transit as against private has become very important recently. The prediction of bus travel and arrival time for the passenger and service provider improves the reliability of the service by reducing the waiting time and uncertainty for the passengers. Several research works have focused on predicting the travel and arrival time of public transit buses in the past. A review of the existing works using parametric and non-parametric machine learning models for bus travel arrival time is conducted in this article. Parameters such as data sources, the spatial-temporal scales, and the study location are considered for the study. It is observed from the review that parametric models are widely used; especially network models, most models use data sourced from the GPS, spatially; segmented routes are well studied, and temporally, all-day data are the most studied. The existing works are well spread across the globe irrespective of developed and developing countries. Overall, it is observed from the survey that bus arrival time prediction is a well-studied area, but still, there is scope for innovative solutions in the future.